...
Note that testing the 6.10 S5R4 App is actually testing the code we will use for S5R5 (in fact the "flaw" I found in the SSE2 App was that it was built as an S5R5 one, i.e. it contained the leap seconds fix that we need to switch off for compatibility to existing S5R4 Apps)....
Your system has been having similar problems for several days now, before Bernd had put up the new SSE2 package, so I don't think it is the "new" SSE2 application...
Is the system overclocked? Have you tested the memory using something like Memtest86? Have you checked for stability using something like Orthos or Prime95? Do you have the most recent drivers for your video card?
Hello Brian
I found that the running of Seti@home beta as a CUDA Task is erroing out my system. When I run S@Hbeta together with E@H 6.10 I had problems. When I run it together with Milkyway I got no problems. So I flushed my cache and will run E@H only and in stock version.
To your questions: Yes it is - Q6600@3,25 temps around 70 degrees (Celsius)
drivers are 185.20 Did not test stability cause in the last 12 month I had not one problem in doing crunching jobs. The errors in E@H were the first ones.
Thank you for the reply
PS: After 15 minutes there was the daily first error in the line. Detached from E@H. Wil get back when it is stable or i have a new rig
I found that the running of Seti@home beta as a CUDA Task is erroing out my system. When I run S@Hbeta together with E@H 6.10 I had problems. When I run it together with Milkyway I got no problems. So I flushed my cache and will run E@H only and in stock version.
I'm not sure what you mean by that. What is the meaning of the word "it" that I bolded and underlined in the quote from you? Do you mean that when you run the CUDA task with Milkyway you don't have problems, or do you mean that when you run E@H with Milkyway you don't have problems?
Also, are you displaying graphics or not? Since you mention a CUDA app being involved, what could be causing the issue is the graphics thread firing up and stepping on some memory. If that's the case, I don't know if it is an issue with E@H or with the CUDA implementation. It would probably be beneficial to work with Bernd on this.
@Archae86, Richard Haselgrove, and/or any other CUDA-capable people: If you're reading this and you participate with SETI and the CUDA app, can any of you test for this phenomenon with your systems? My poor (and old) 6800GT is not CUDA-compliant...
@Archae86, Richard Haselgrove, and/or any other CUDA-capable people: If you're reading this and you participate with SETI and the CUDA app, can any of you test for this phenomenon with your systems? My poor (and old) 6800GT is not CUDA-compliant...
Sorry, not me - I've got no fancy graphics cards at all. I did consider installing one to participate in the Beta testing, but I lost interest when the SETI application was declared fit for live use with no further development.
However, I think your theory is a plausible one: and if I remember rightly, Einstein graphics can be disabled by placing a dummy null file with an appropriate name in the BOINC directory. I'll go and search for the reference: maybe useful for someone to test with.
Edit - that name was EAH_NO_GRAPHICS: description of usage here
@Archae86, Richard Haselgrove, and/or any other CUDA-capable people: If you're reading this and you participate with SETI and the CUDA app, can any of you test for this phenomenon with your systems? My poor (and old) 6800GT is not CUDA-compliant...
I do run SETI-CUDA and Einstein but I don't remember if I've tried the Einstein-graphics at the same time as a cuda-task has been running...
Can try when I get home from work in about 2 hours.
@Archae86, Richard Haselgrove, and/or any other CUDA-capable people: If you're reading this and you participate with SETI and the CUDA app, can any of you test for this phenomenon with your systems? My poor (and old) 6800GT is not CUDA-compliant...
Sorry, not me - I've got no fancy graphics cards at all. I did consider installing one to participate in the Beta testing, but I lost interest when the SETI application was declared fit for live use with no further development.
You'll have that... Similar to AP and AMD, but I digress...
Quote:
However, I think your theory is a plausible one: and if I remember rightly, Einstein graphics can be disabled by placing a dummy null file with an appropriate name in the BOINC directory. I'll go and search for the reference: maybe useful for someone to test with.
Edit - that name was EAH_NO_GRAPHICS: description of usage here
Yeah, I had forgotten about that flag file. I don't mind the star sphere, so I just let it run... Of course my Intel P4 system is a service install, so no graphics on that anyway. I hope Sabroe gives it a shot to see if that helps...
I just made this App "official". This gives you the opportunity now to switch back to the "official" path (you should empty your work cache before removing the app_info.xml), which I would recommend. This will allow you to get ABP1 and S5R5 work right away when we issue it.
However, I think your theory is a plausible one: and if I remember rightly, Einstein graphics can be disabled by placing a dummy null file with an appropriate name in the BOINC directory. I'll go and search for the reference: maybe useful for someone to test with.
Edit - that name was EAH_NO_GRAPHICS: description of usage here
Yeah, I had forgotten about that flag file. I don't mind the star sphere, so I just let it run... Of course my Intel P4 system is a service install, so no graphics on that anyway. I hope Sabroe gives it a shot to see if that helps...
This file has no effect since we are using "APIv6" Apps, which run the graphics in a separate program instead of a thread. Just avoid the graphics program to be started, i.e. by choosing a different screen saver than "BOINC".
In the 6.10 Beta package you could remove the tags referring to the *_graphics_* files from the app_info.xml if you want to disable the graphics just for Einstein@home.
@Archae86, Richard Haselgrove, and/or any other CUDA-capable people: If you're reading this and you participate with SETI and the CUDA app, can any of you test for this phenomenon with your systems? My poor (and old) 6800GT is not CUDA-compliant...
I do run SETI-CUDA and Einstein but I don't remember if I've tried the Einstein-graphics at the same time as a cuda-task has been running...
Can try when I get home from work in about 2 hours.
Well, home from work now and Boinc is running 2 Einstein tasks, 1 Climate prediction task and a Seti-Cuda. No problems. Did try graphics on both Einstein and Climate and it works like a charm, have to admit that it lags a bit now and then but I suppose that's because my GPU is running Seti at the same time.
Have been lucky and only got 2 errors running Cuda, and that's probably the Cuda-apps fault!
System:
Intel Q9450, Win Vista home premium 64-bit
GeForce 9600GT 512Mb, temp @40-41C under load.
4 Gb RAM
All at stock speed.
I think any problem observed that seems to coincide with CUDA apps and E@H app running at the same time are most likely caused by thermal problems for overclocked CPUs. The current E@H apps are stressing the CPU pretty well (someone reported that using the latest beta (now official) app resulted in a 6°C rise in core temp. of his PC). Add a GPU that is running under full load in the same computer case, plus overclocking ... that's quite heavy on the cooling.
I think any problem observed that seems to coincide with CUDA apps and E@H app running at the same time are most likely caused by thermal problems for overclocked CPUs. The current E@H apps are stressing the CPU pretty well (someone reported that using the latest beta (now official) app resulted in a 6°C rise in core temp. of his PC). Add a GPU that is running under full load in the same computer case, plus overclocking ... that's quite heavy on the cooling.
I didn't, and still don't, know what proper maximum temps for current CPUs should be, so I didn't comment on the 70C temperature before. My gut feeling though is it is a bit high. My suggestion would probably be to back down on the overclocking to in/around the 3GHz range instead of 3.25 and see if that helps...
One other thing it could be is memory being overclocked too aggressively or with too tight timings, so slowing the memory down or loosening up the timings might help...
I'd still suggest anyone who can run the CUDA application test them running simultaneously. It is going to become more the norm than the exception that people are doing that, so if there is a correlation, it would be good to find it sooner than later...
RE: RE: RE: ... Note
)
Hello Brian
I found that the running of Seti@home beta as a CUDA Task is erroing out my system. When I run S@Hbeta together with E@H 6.10 I had problems. When I run it together with Milkyway I got no problems. So I flushed my cache and will run E@H only and in stock version.
To your questions: Yes it is - Q6600@3,25 temps around 70 degrees (Celsius)
drivers are 185.20 Did not test stability cause in the last 12 month I had not one problem in doing crunching jobs. The errors in E@H were the first ones.
Thank you for the reply
PS: After 15 minutes there was the daily first error in the line. Detached from E@H. Wil get back when it is stable or i have a new rig
RE: I found that the
)
I'm not sure what you mean by that. What is the meaning of the word "it" that I bolded and underlined in the quote from you? Do you mean that when you run the CUDA task with Milkyway you don't have problems, or do you mean that when you run E@H with Milkyway you don't have problems?
Also, are you displaying graphics or not? Since you mention a CUDA app being involved, what could be causing the issue is the graphics thread firing up and stepping on some memory. If that's the case, I don't know if it is an issue with E@H or with the CUDA implementation. It would probably be beneficial to work with Bernd on this.
@Archae86, Richard Haselgrove, and/or any other CUDA-capable people: If you're reading this and you participate with SETI and the CUDA app, can any of you test for this phenomenon with your systems? My poor (and old) 6800GT is not CUDA-compliant...
RE: @Archae86, Richard
)
Sorry, not me - I've got no fancy graphics cards at all. I did consider installing one to participate in the Beta testing, but I lost interest when the SETI application was declared fit for live use with no further development.
However, I think your theory is a plausible one: and if I remember rightly, Einstein graphics can be disabled by placing a dummy null file with an appropriate name in the BOINC directory. I'll go and search for the reference: maybe useful for someone to test with.
Edit - that name was EAH_NO_GRAPHICS: description of usage here
RE: @Archae86, Richard
)
I do run SETI-CUDA and Einstein but I don't remember if I've tried the Einstein-graphics at the same time as a cuda-task has been running...
Can try when I get home from work in about 2 hours.
RE: RE: @Archae86,
)
You'll have that... Similar to AP and AMD, but I digress...
Yeah, I had forgotten about that flag file. I don't mind the star sphere, so I just let it run... Of course my Intel P4 system is a service install, so no graphics on that anyway. I hope Sabroe gives it a shot to see if that helps...
I just made this App
)
I just made this App "official". This gives you the opportunity now to switch back to the "official" path (you should empty your work cache before removing the app_info.xml), which I would recommend. This will allow you to get ABP1 and S5R5 work right away when we issue it.
BM
BM
RE: RE: However, I think
)
This file has no effect since we are using "APIv6" Apps, which run the graphics in a separate program instead of a thread. Just avoid the graphics program to be started, i.e. by choosing a different screen saver than "BOINC".
In the 6.10 Beta package you could remove the tags referring to the *_graphics_* files from the app_info.xml if you want to disable the graphics just for Einstein@home.
BM
BM
RE: RE: @Archae86,
)
Well, home from work now and Boinc is running 2 Einstein tasks, 1 Climate prediction task and a Seti-Cuda. No problems. Did try graphics on both Einstein and Climate and it works like a charm, have to admit that it lags a bit now and then but I suppose that's because my GPU is running Seti at the same time.
Have been lucky and only got 2 errors running Cuda, and that's probably the Cuda-apps fault!
System:
Intel Q9450, Win Vista home premium 64-bit
GeForce 9600GT 512Mb, temp @40-41C under load.
4 Gb RAM
All at stock speed.
Hi! I think any problem
)
Hi!
I think any problem observed that seems to coincide with CUDA apps and E@H app running at the same time are most likely caused by thermal problems for overclocked CPUs. The current E@H apps are stressing the CPU pretty well (someone reported that using the latest beta (now official) app resulted in a 6°C rise in core temp. of his PC). Add a GPU that is running under full load in the same computer case, plus overclocking ... that's quite heavy on the cooling.
CU
Bikeman
RE: I think any problem
)
I didn't, and still don't, know what proper maximum temps for current CPUs should be, so I didn't comment on the 70C temperature before. My gut feeling though is it is a bit high. My suggestion would probably be to back down on the overclocking to in/around the 3GHz range instead of 3.25 and see if that helps...
One other thing it could be is memory being overclocked too aggressively or with too tight timings, so slowing the memory down or loosening up the timings might help...
I'd still suggest anyone who can run the CUDA application test them running simultaneously. It is going to become more the norm than the exception that people are doing that, so if there is a correlation, it would be good to find it sooner than later...